UTICalc 3.0 demonstrated strong accuracy for predicting urinary tract infections in febrile pediatric patients, according to a prospective diagnostic study published in JAMA Network Open.
Researchers enrolled 2,561 patients aged 2 to 24 months with a fever (temperature of 100.4 °F [38 °C] or higher) at two Canadian pediatric emergency departments from November 2022 to January 2025. Urinary tract infections (UTIs) were classified in 111 patients, including cases meeting predefined urinalysis and culture criteria as well as a small number identified on follow-up or adjudication.
Researchers evaluated two UTICalc models using predefined risk thresholds. The clinical model incorporated five variables: age younger than 12 months, sex and circumcision status, temperature greater than 39 °C, duration of fever greater than 48 hours, and absence of another source of fever. The clinical and dipstick model additionally included leukocyte esterase and nitrite results from urinalysis.
The clinical model achieved an area under the receiver operating characteristic curve (AUROC) of 84%. At the 2% risk threshold, sensitivity was 96% and specificity was 34%. At the 5% threshold, sensitivity decreased to 82% while specificity increased to 74%.
The clinical and dipstick model had an AUROC of 95%, with 94% sensitivity and 87% specificity at the 5% threshold. The model was not evaluated at the 2% threshold.
The clinical model’s recommendations for urine testing were also compared with those of emergency department physicians. Physicians tested 1,157 patients and missed two UTI cases (sensitivity 98%, specificity 57%). The clinical model recommended testing 1,719 patients (missing four UTIs) at the 2% threshold and 731 patients (missing 20 UTIs) at the 5% threshold.
Decision curve analysis showed a positive net benefit for both UTICalc models across a range of thresholds. The clinical model demonstrated the greatest clinical benefit between 2% and 6% thresholds, while the clinical and dipstick model showed the greatest benefit between 4% and 12%.
Limitations included missing urine test results, incomplete patient follow-up, reliance on caregiver-reported information, and the possibility of missed UTIs among patients treated for other conditions.
The findings suggest both models may support clinical decision-making by identifying pediatric patients more likely to benefit from urine testing, although UTICalc did not outperform experienced clinicians in the tertiary pediatric emergency departments studied.
“Physicians should still rely on their clinical judgement when deciding to obtain urine tests, but the UTICalc models can serve as a guide in situations when the diagnosis may not be clear,” said lead study researcher Ceilidh Kinlin, MD, of the Department of Pediatrics, Children’s Hospital of Eastern Ontario, at the University of Ottawa.
One researcher reported serving as a minority shareholder in a concussion care network, and another reported institutional grant funding. No other conflicts were disclosed.
Source: JAMA Network Open
